JP3006560B2 - Alignment device and computer-readable recording medium storing alignment program - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a positioning apparatus for automatically obtaining a positional relationship between partial images from the characteristics of the partial images in order to input a wide range of images to the computer. It relates to a readable recording medium.

[0002]

2. Description of the Related Art Conventionally, electronic cameras and scanners are generally used to input images into a computer.

However, since the angle of view and the reading range are limited in both electronic cameras and scanners, it may not be possible to read an object over a wide area at one time. In such a case, it is necessary to perform a process of reconstructing a target image obtained by dividing into a plurality of images. As a technique for achieving such an object, a positioning apparatus that performs processing for automatically applying positioning from the characteristics of an image by a computer is known as a conventional technique.

[0004] In this conventional positioning apparatus, parameters are often found by successively searching in a parameter space formed from all parameter candidates using an index of a residual or a cross-correlation coefficient. These methods are described in "Kyoritsu Shuppan, bit separate volume" Computer Science ", ACM Computing Surveys '92,
77-119 ", the details of which are omitted, but there is a problem that the calculation amount increases exponentially as the types of fluctuations included increase. To avoid the problem of the calculation amount, for example, Japanese Patent Application Laid-Open No. 7-264483,
“Radio image alignment method” has been proposed. In this proposal, first, a reference feature point is extracted from an image, and an image around the feature point is regarded as a block. Then, only the process of obtaining the translation amount (block matching process) is applied to each block, the results are integrated by the least squares method, and the parameters including the translation amount, the rotation amount, and the enlargement / reduction amount are batched. It describes the method of determining In this method, since the parameter space is constituted only by the amount of translation, it is possible to prevent an exponential increase in the amount of calculation. Also,
"Japanese Patent Laid-Open No. 8-145628," Method of Locating Object ""
In the method, a parameter including a translation amount and a rotation amount for all combinations is obtained from a result of applying the block matching process to a plurality of templates, and a parameter that gives a minimum square error among the parameters is used as an alignment result. Described.

[0005]

However, Japanese Patent Application Laid-Open No. Hei 7-264483, "Method of Aligning Radiation Images"
In the conventional alignment method based on block division such as, the parameter space is composed of only the amount of translation, it is possible to prevent an exponential increase in the amount of computation, but the obtained parameter is incorrectly translated. There is a problem that the block may give an influence to the amount, and the accuracy of the obtained parameter is low.

In the positioning method according to "Japanese Patent Application Laid-Open No. Hei 8-145628," Method of Locating Object ", a parameter for minimizing a square error is obtained from parameters obtained for every combination of blocks. By that
We are trying to avoid the effects of blocks that give incorrect translations, but because the optimal parameters are determined based only on the block matching processing results, multiple blocks may have similar incorrect translations. Giving them did not always give the correct result.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem, and has as its object to provide a positioning apparatus for determining a positional relationship between images including rotation at high speed and with high accuracy.

[0008]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, in a positioning apparatus according to the present invention, when a positional relationship between images including rotation is obtained by integrating block matching processing results, an erroneous position is obtained. Block selecting means for eliminating blocks likely to produce block matching processing results, matching result classifying means for classifying the obtained block matching processing results, and a parameter representing a position correction amount of the target image for each classified group , And an optimum position correction parameter determining unit that determines a position correction parameter that gives the best result by referring to the original image from the plurality of obtained position correction parameters. .

Therefore, in the positioning apparatus according to the present invention, since the block which has a high possibility of giving an erroneous block matching processing result by the block selecting means is excluded, the accuracy of the positioning result of each block obtained by the block matching processing is improved. improves. Furthermore, the block matching processing result is obtained by the matching result classifying means,
The same position correction parameters are classified into groups that are considered to be generated, and the accuracy of the position correction parameters obtained by obtaining the parameters by the position correction parameter calculation means for each group is improved. In addition, since the optimal parameter is determined by the optimal parameter determining means with reference to the original image from the plurality of obtained position correction parameters, the amount of calculation can be reduced and the positioning can be performed with high accuracy.

[0010]

Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows an alignment device 40 according to the present invention.
FIG. 4 is a block diagram illustrating an example of a configuration of a zero.

Referring to FIG. 1, a positioning device 400 includes:
Divide the reference image that serves as the alignment reference into blocks,
The attribute information of each of the divided blocks and the position parameter candidates at the time of applying the registration are obtained, and the block attribute
7, the characteristic value of the block is obtained from the attribute information in the block attribute memory 407 and the image information in the image memory 108, and there is little possibility that an erroneous result will occur when the alignment is applied. Block selecting means 402 for selecting only blocks determined to be
Each block selected by the block selecting unit 402 is obtained by the block matching unit 403 that applies the alignment with reference to the reference image in the image memory 108 and the information of the target image to which the alignment is applied. The matching result classifying means 404 for classifying blocks regarded as having originated from the same position correction parameter into the same group from the alignment results of the respective blocks, and the position correction parameter is obtained for each group classified by the matching result classifying means 404. From the position correction parameter calculation means 405 stored in the parameter memory 408 and the position correction parameters for each group calculated by the position parameter calculation means 405 and stored in the position parameter memory 408, the reference image and the target image have the best weight. Ri optimal position correction parameter determining means 406 for selecting a position correction parameter of the state, the block attribute memory 407 for storing the position parameter candidates to be used for the attribute information and the matching of each block,
A position parameter memory 408 for storing position parameter candidates of the target image and position correction parameters obtained for each group, a reference image serving as a reference for positioning, and partial images such as a target image to which positioning is applied. And an image memory 108 for storing each image.

Next, an outline of the operation of the present invention will be described.

In the registration apparatus 400 according to the present invention, first, a reference image serving as a reference for registration in the image memory 108 is divided into blocks by the block dividing means 401. Attribute information such as the position of each divided block is stored in the block attribute memory 407 together with a position parameter candidate group indicating the applicable range of the alignment. Based on the attribute information in the block attribute memory 407, a characteristic value for evaluating the possibility of generating an erroneous result when the block matching processing is applied is calculated for each block by the block selecting unit 402, and the erroneous result is calculated. Only the blocks that are unlikely to generate are selected. The blocks selected by the block selection unit 402 include:
The block matching unit 403 applies a positioning process based on the parallel movement, determines a point corresponding to the reference image and the target image for each block, and
7 is stored. The matching result classifying means 404 classifies the result of the positioning process for each block into groups based on whether or not they are based on the same position correction parameter.
The position correction parameter calculation unit 405 calculates a position correction parameter including the amount of rotation for each group classified by the matching result classification unit 404. From the obtained position correction parameters including the amount of rotation for each group, a position correction parameter that best overlaps the reference image and the target image is determined by the optimum position correction parameter determination unit 406 with reference to the reference image and the partial image. .

[0015]

Next, specific embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 illustrates an example of the processing flow of the positioning apparatus 400 of the present invention, and FIG.
Each configuration of the present invention shown in FIG.

In the following description, an image serving as a reference for positioning is referred to as a reference image 451, and an image serving as a target of positioning processing is referred to as a target image 452. It is assumed that the initial position of the target image 452 in the reference image coordinate system 450, a set of possible position correction parameter candidates, and image attributes are stored in the position parameter memory 408 in advance. These can also be configured to be read from the outside by communication means. The details of this configuration are described in Japanese Patent Application Laid-Open No. Hei 9-218941, entitled "Image Input Apparatus", which was previously filed by the present applicant, and a description thereof will be omitted. The initial position of the target image 452 is represented by (X, Y, θ) (Equation 1) in a reference image coordinate system 450 having the origin at the center of the reference image as shown in FIG. At this time, the position (M, N) in the reference image coordinate system 450 corresponding to the position (m, n) in the target image coordinate system 460 is expressed by using a homogeneous expression.

[0018]

(Equation 1)

## EQU1 ##

It is assumed that only fluctuations due to rotation and translation are included between images, and that no enlargement / reduction or skew is included. The size of the image is set to (Sx, Sy) for both the reference image 451 and the target image 452, but these are set for the sake of simplicity. It is not limited to. When the position of the target image 452 is expressed by a method different from that of the present embodiment, or when the position includes a more complicated change such as enlargement / reduction, the reference image 451 and the target image 4
The present invention is also applicable to the case where the sizes of 52 are different.

In the positioning apparatus 400, first, the reference image 451 is divided into blocks of a fixed size (Bx, By) by the block dividing means 401, and information of each block is stored in the block attribute memory 407. The size of the block is determined by the content and size of the image, but is usually 16 × for an image of about 320 × 240 pixels.
A size of about 16 or 8 × 8 is sufficient. Each block is numbered sequentially from 1 without duplication, and the i-th (1 ≦ i ≦ N) block is referred to as a block i. This is set for convenience of explanation, and may be any type as long as each block can be designated. The information of the block i is the center position 454 (x
i, yi), a flag Fi indicating whether or not to apply the block matching process, and a group number Gi indicating the classification result by the matching result classification unit 404. F
When i is 0, the block matching process is not applied to block i. The group number Gi may be any number as long as the group number can be identified, and may be used also as the matching application flag Fi. In this case, 0 when no block matching processing is applied,
A block matching process may be applied, and if the data is classified into a group, the group number may be given. In addition, information indicating the characteristics of the block can be added. At the same time, the block dividing means 4
01 sets a position parameter candidate that can be taken by the block i from the position parameter candidate group stored in the position parameter memory 408 and stores it in the block attribute memory 407. It is also possible to configure so that all blocks have the same position parameter candidate.

Next, the positioning device 400 selects a block to which the block matching process is applied by the block selecting means 4.
02 to select. This is because block i is the reference image 4
By setting the value of the block-matching application flag Fi to 1, when it is included in the overlapping portion of the target image 452 and the target image 452 at the initial position and it is unlikely that an incorrect result will be obtained when the block matching process is applied. carry out. The former can be determined from the initial positional relationship between the reference image 451 and the target image 452, and the latter can be determined from the reference image 451 based on the characteristic value closely related to the result of the block matching process.

First, a method for determining whether or not a block is included in an overlapping portion at the initial position between the reference image 451 and the target image 452 will be described.

The position (a,
The position (a ′, b ′) expressing b) in the target image coordinate system 460 is a ′ = (a−X) × cos (−θ) − (by) × sin (−θ) (Equation 3) -1) b ′ = (a−X) × sin (−θ) + (by) × cos (−θ) (Formula 3-2) The positions of the four corners of the block i in the reference image coordinate system 450 are: upper left (xi-Bx ÷ 2, yi-By ÷ 2) (Equation 4-1) lower left (xi-Bx ÷ 2, yi + By ÷ 2) (Equation 4) -2) Lower right (xi + Bx ÷ 2, yi + By ÷ 2) (Equation 4-3) Upper right (xi + Bx ÷ 2, yi-By ÷ 2) It is obtained as (Equation 4-4). When the results of expressing the positions of the four corners in the target image coordinate system 460 by Expression 3 are all on the target image 452, it can be considered that the block i is completely included in the overlapping portion. Although the reference image 451 and the target image 452 have been described as being rectangular here, the present invention can also be applied to an image having a more complicated shape using a mask image.
It may be configured to determine whether or not the position of the corner coordinates in the target image coordinate system 460 is on the target image 452 with reference to the mask image.

Next, a description will be given of a method of judging a possibility that an erroneous result is obtained when the block matching processing is applied.

In this embodiment, the variance of the block i is obtained with reference to the reference image 451. If the variance is equal to or smaller than a predetermined threshold value, an erroneous result is obtained even if the block matching processing is applied to the block i. to decide. FIG.
Shows an example of a block and its density distribution. In the block B462, it can be seen that the pixel values are concentrated at a constant value, and that there are few pixels effective for matching. Since good results cannot be obtained by applying matching to such a block, the block selecting unit 402 determines such a pixel from the variance of the density distribution and excludes such a pixel from the selection. Various characteristic values other than the variance can be used. 2
In the value image, the ratio of the black pixels included in the block can be used. "Yoshida et al.," Reliability Functions for Motion Vectors of Moving Images and Their Applications, "IEICE Transactions D-
II, Vol. J80-D-II, No. 5, pp. 1192-1201, 1997 ”, it is also possible to use more complicated characteristic values. FIG. 5 is a diagram showing an example of block selection, in which a block not shaded is a selected block.

In the above description, all blocks that are included in the overlapping portion of the reference image 451 and the target image 452 at the initial position and that are unlikely to lead to an erroneous result when the block matching process is applied are selected. But
It is not always necessary to select all blocks that satisfy the conditions. It is also possible to configure so that only a predetermined number of blocks are selected in the descending order of the characteristic values of the blocks such as variance. At this time, the number of blocks to be subjected to the block matching processing can be limited, so that the processing amount required for the alignment processing can be further reduced.

If the number of blocks selected by the block selecting means 402 is two or more, steps F04 to F07
The block matching process is applied to the block selected by the above, and the position correction parameter of the target image 452 is determined (step F03). If the number of selected blocks is less than 2, the parameters including the rotation cannot be determined from the block matching processing, and the processing ends. At this time, it is also possible to adopt a configuration in which the position correction parameter is obtained by using a conventional technique such as a method of performing a sequential search in a parameter space formed from all the position parameter candidates, instead of simply completing the process. . In such a configuration, since the sequential search in the parameter space is performed only when the positioning apparatus according to the present invention cannot be applied, when performing positioning processing a plurality of times, such as when applying positioning between a plurality of images, a conventional method is used. The processing efficiency can be improved as compared with the case where the full search in the parameter space is simply applied.

The selected block is subjected to block matching processing by the block matching means 403 (step F04).

FIG. 6 is a diagram for explaining the block matching process.

The block matching process is performed at block i45.
In order to determine the position on the target image 452 corresponding to 3,
By referring to the block attribute memory 407 storing the position parameter candidate group obtained by the block dividing unit 401, the translation amount component of the target image 452 used for matching is obtained, and the search range 456 which is the movement range of the block i 453 is determined. Set. Center position 4 of block i453
54 is moved within the search range 456, the degree of overlap between the block i 453 and the target image 452 at each position is evaluated by a similarity measure, and the position giving the highest similarity measure is determined. This position is regarded as the position where the block i 453 and the target image 452 have the highest similarity. This is the basic image processing used as template matching. The details are omitted since they are introduced in detail in “Image Processing Handbook”, pp. 303-305 ”and“ The University of Tokyo Press, “Image Analysis Handbook”, pp. 707-746 ”. In addition, in alignment with an image having a complicated shape or a rotated image using a mask image, a block i453
In some cases, the shape of the target image 452 corresponding to is not rectangular, but even in such a case, the registration method according to “Japanese Patent Application No. 9-56060,“ Image Registration Method ”” previously proposed by the inventor. Is applied, the block matching process can be performed at high speed.

Hereinafter, as a result of the block matching processing in step F04, a block i 453 at a position 454 (xi, yi) and a point 455 (x
i, yi) correspond to each other.

In the positioning apparatus according to the present invention, when block matching processing is applied by the block selecting means 402, blocks which are likely to lead to erroneous results are excluded. , There is a case where a matching result different from the actual image variation due to the characteristics of the image or an erroneous matching result due to an influence of an error is given. In order to eliminate such an influence, the results obtained by the block matching processing are classified, and those having a high possibility of generating the same position correction parameter are grouped into the same group (step F05). The correction parameters are determined (step F06).

FIG. 7A is a diagram for explaining a method of classifying blocks, and shows the relationship between block i and block j.

The positions of blocks i and j are respectively (xi,
yi), (xj, yj), the positions of points on the target image 452 corresponding to the blocks i, j are (Xi, Yi),
(Xj, Yj), the vector from the block i to the block j is (dx, dy), and the target image 4 corresponding to the block i
Let a vector from a point on 52 to a point on the target image 452 corresponding to the block j be (dX, dY). Target image 4
The position correction parameter for moving the position 52 to the position that best overlaps the reference image 451 is represented by a rotation amount θ ′ and a parallel movement amount (X ′, Y ′) using a homogeneous expression.

[0036]

(Equation 2)

Can be written as

When the block i and the block j and the matching result thereof are in accordance with Equation 5, the transformation for moving the position of the point on the target image 452 corresponding to the block i to the block i is as follows.

[0039]

(Equation 3)

The conversion for moving the position of the point on the target image 452 corresponding to the block j to the block j is as follows.

[0041]

(Equation 4)

And the difference (dx, dy) between the two is

[0043]

(Equation 5)

Is as follows. From the above equation 8, when the same transformation is applied to blocks i and j, (dx, dy), (dX,
dY) exists at a position separated by the rotation amount θ ′ on the same circumference as shown in FIG. 7B. If the range of the rotation amount θ ′ and the error of the allowable distance can be limited, the range of (dX, dY) relative to (dx, dy) can also be limited. Therefore, whether (dX, dY) obtained from the block matching processing result is included in this existence range is determined as to whether or not the result of applying the block matching processing to the blocks i and j is based on the same conversion.

The range of the rotation amount θ ′ is given in a position correction parameter candidate in the position parameter memory 408. Here, it is assumed that θ ′ is in the range of −Δθ ≦ θ ′ ≦ Δθ (Δθ ≧ 0) (Equation 9) with 0 as the center. Also, (dx, dy), (d
X, dY) and the allowable range of the difference d between the origin and the origin are given by -Δr ≦ d ≦ Δr (Δr ≧ 0) (Equation 10). These are set for the purpose of explanation, and are determined depending on the object to which the positioning device of the present invention is applied.

At this time, (dx, dy) and (dX, dY)
The angle dθ and the distance difference dr are

[0047]

(Equation 6)

Is as follows. If dθ and dr satisfy −Δθ ≦ dθ ≦ Δθ (Δθ ≧ 0) (Equation 12) and −Δr ≦ dr ≦ Δr (Δr ≧ 0) (Equation 13), block matching processing is performed on blocks i and j. It is considered that the result obtained by applying is based on the same position correction parameter, and both are considered to be in the same group.

By applying this operation to all combinations for selecting two blocks, the whole is grouped, and the blocks belonging to the same group have the same value as the group number Gi of the block attribute information in the block attribute memory 407. Given. As a result, the block matching processing results are classified into groups that provide the same position correction parameter. In this process, blocks that do not belong to any group are block matching processing results that are clearly different from other blocks, and matching results that differ from actual image fluctuations due to image characteristics, and incorrect matching due to errors. It is believed that results have been produced. The accuracy of the position correction parameters obtained by eliminating such blocks and classifying the blocks into groups that provide the same position correction parameters is improved. FIG. 8 shows an example of classification by the matching result classifying means 404. The block matching processing result (FIG. 8A) of each block is classified into two groups and excluded blocks (FIG. 8A). FIG. 8 (b).

In the above description, the similarity between the blocks i and j is determined in two stages as to whether or not dθ and dr are within a certain range.

[0051]

(Equation 7)

It is also possible to make a determination using an evaluation function as described above. When the evaluation function exceeds the set threshold value, it is considered that the block matching processing results of the blocks i and j are generated from the same position correction parameter. How to set the evaluation function and the threshold value needs to be determined in consideration of the characteristics of the image. However, compared with the above two-stage evaluation, it is possible to determine a block existing near the boundary of the existence range. The feature is that it can be performed flexibly.

From each of the groups thus obtained,
The position correction parameter calculation unit 405 obtains a position correction parameter for correcting the position of the target image 452 using the least squares method (Step F06).

That is, when the position correction parameter of the target image 452 is expressed by the above equation 5, θ ′,
x ′ and y ′ are the target images 4 corresponding to the block i453
The value obtained by applying the position correction parameter to the position 455 (Xi, Yi) on the position 52 and the position 454 of the block i 453
(Xi, yi),

[0055]

(Equation 8)

The value obtained by adding the square of the size of the blocks belonging to the same group to the minimum is minimized, that is,

[0057]

(Equation 9)

Θ ′, x ′, and y ′ that satisfy the above conditions. this is,

[0059]

(Equation 10)

Is obtained by solving the simultaneous equations

[0061]

[Equation 11]

Is obtained. Here, Σ indicates the sum of all blocks included in the same group, and the number of blocks included in the same group is K.

The position correction parameters thus obtained are stored in the position parameter memory 408. This position correction parameter indicates the conversion from the initial position of the target image 452 in the reference image coordinate system 450.

[0064]

(Equation 12)

By multiplying with the reference image coordinate system 4
A new position of the target image 452 at 50 is obtained.

In steps F05 to F06, the optimum position correction parameter determining means 406 is determined from the plurality of position correction parameters stored in the position parameter memory 408.
As a result, the position correction parameter for superimposing the target image 452 on the reference image 451 in the best manner is determined (step F07). That is, the similarity of the overlapping portion between the reference image 451 and the target image 452 when the target image 452 is moved by sequentially applying the position correction parameters to the initial position of the target image 452 is evaluated, and the position giving the best similarity is evaluated. The correction parameter is set as an optimum position correction parameter. This is a positioning process by sequential search using a plurality of position correction parameters in the position parameter memory 408 as a parameter candidate group, and a conventional positioning method can be used. As a matter of course, it is also possible to apply an alignment method based on “Japanese Patent Application No. 9-56060,“ Image alignment method ”” filed by the present applicant before filing the present invention.

As described above, the position correction parameters determined by the optimum position correction parameter determining unit 406 indicate conversion of the target image 452 to a position where the target image 452 can be superimposed on the reference image 451 best. It is given as a result of the alignment processing of the target image 452.

As described above, in the positioning apparatus according to the present invention, the positioning apparatus for obtaining the best superposition position of the target image 452 with respect to the reference image 451 from the result of the block matching processing for the parallel movement is used as a block. By selecting only blocks that are likely to produce correct results when the block matching process is applied by the selection unit 402 and classifying and grouping the block matching process application results to the selected blocks, an erroneous result is obtained. Blocks having a high possibility of obtaining a result are eliminated, the accuracy of the position correction parameters including the rotation obtained for each group by the position correction parameter calculation means 405 is improved, and among the position correction parameters obtained for each group, The reference image 451 is obtained by the optimal position correction parameter determining unit 406. Determining a position correction parameter target image 452 overlaps the best. With such a configuration, it is possible to obtain the best superposition position of the target image 452 with respect to the reference image 451 at high speed and with high accuracy.

In the above description, the shapes of the blocks are all the same size and are generated by mechanically dividing the reference image 451, but can be used as a feature for providing a clue for positioning. An edge or the like may be extracted first, and a block may be set near the feature point. In such a configuration, the block dividing process in the block dividing unit 401 and the process of selecting only blocks that are likely to generate a correct result when the correct block matching process in the block selecting unit 402 is applied are performed at the same time. Can be considered
Even in such a configuration, by classifying and applying the block matching processing application results, blocks having a high possibility of obtaining an erroneous result are eliminated, and the rotation obtained for each group by the position correction parameter calculation unit 405 is determined. The accuracy of the included position correction parameters is improved, and among the position correction parameters obtained for each group, the reference image 451 and the target image 452 are most determined by the optimal position correction parameter determining unit 406 by the optimal position correction parameter determining unit 406. Since the position correction parameters that satisfactorily overlap are determined, the reference image 451 of the target image 452 of the present invention is used.
The effect of being able to obtain the best superposition position with high speed and high accuracy is not lost.

The image position adjusting method and the position aligning device according to the present invention are one of the basic processing devices required when configuring an apparatus for handling images on a computer. Applicable.

An example will be described below.

FIG. 9 is a diagram showing an example of the configuration of a composite image generating apparatus 100 using the positioning apparatus of the present invention.

Referring to FIG. 9, a composite image generating apparatus 100
Is an image position adjustment device 300 that performs image position adjustment,
Position correcting means 200 configured by using the positioning device 400 according to the present invention, an image memory 108 for storing a partial image, and the position and size of the partial image stored in the image memory 108 in the composite image coordinate system , An image selection button 312 for transmitting an image selection instruction from the user to the image position adjustment device 300, an image to be position-adjusted, and an amount of movement of the selected image Pointing means 313 used by the image position adjusting device 300 to determine the direction and direction; a display switching button 314 for instructing the image position adjusting device 300 to allow the user to switch the content of the display image; A display unit 601 for displaying an output result of the apparatus 300; An image combining unit 107 for combining in accordance with the information in the memory 109, the partial image from the outside to the image memory 108, the partial image picture attribute memory 10 the attribute information of the
9 and a communication unit 600 for transmitting the synthesized image obtained by the image synthesis unit 107 to an external device.
And

FIG. 10 is a diagram showing an example of the configuration of the position correcting means 200 in the image synthesizing apparatus 100.

In FIG. 10, the position correcting means 200
The positioning device 400 according to the present invention, a positioning image selection unit 201 for selecting a reference image and a target image and transmitting the selected image to the positioning device 400, and obtaining the position of the target image viewed from the reference image and transmitting the position to the positioning device 400 At the same time, the position parameter converting means 20 for obtaining the position of the target image after the positioning from the output result of the positioning device 400
And 2.

In the composite image generating apparatus 100, first, the user transmits the partial image and its attribute information from the outside through the communication means 600 to the image memory 108 and the image attribute memory 10
Read in 9. It is assumed that the imaging numbers are added to the partial images in the reading order. This may be any means used to identify each partial image. The read partial image is displayed on the display unit 601 by the image position adjusting device 300. The user adjusts the image position using the image position adjustment device 300 so that the positional relationship between the partial images displayed on the display unit 601 is correct. After the manual image position adjustment is completed, the target image i and the reference image j corresponding to the target image are selected by the alignment image selection unit 201, and the position of the target image i is set by the position parameter conversion unit 202 to the reference image coordinates. It is converted into a system and transmitted to the positioning device 400.
The positioning device 400 performs positioning based on the given information, and returns a position correction parameter for correcting the position of the target image i to a position where the reference image j and the target image i are superimposed best, Position parameter conversion means 2
02 is the target image i using this position correction parameter.
Is obtained in the new reference image image coordinate system, the obtained position is converted again to the composite image coordinate system, and the position of the target image i in the image attribute memory 109 is updated.

As a method of selecting the reference image j by the alignment image selecting means 201, a method of using a partial image having the imaging number i-1 immediately before the target image as a reference image, and the area of an overlapping portion with the target image is the smallest. There is a method of using a large partial image as a reference image. After the position correction processing of all the partial images by the position correction unit 200 is completed, the image in the image memory 108 is synthesized by the image synthesis unit 107 according to the attribute information in the image attribute memory 109 and stored in the image memory 108. You. The generated composite image is sent to an external device by the communication unit 600.

As described above, in the composite image generating apparatus 100 using the positioning apparatus 400 according to the present invention, since the precise positioning is executed by the positioning apparatus 400, the user can exactly adjust the position between the images. Does not need to be performed, and it is only necessary to roughly adjust the position while referring to the feature amount of the image. Therefore, high-accuracy positioning can be performed easily and at high speed.

The applicant has filed a patent application (Japanese Patent Application No.
-26350; Japanese Patent Application Laid-Open No. 9-218941), an image in which a plurality of partial images of a target imaged by an imaging unit are obtained by moving the imaging unit, and a synthesized image is generated from the obtained partial image group by an image synthesis process. Although an input device has been proposed, the alignment device of the present invention can be used as a device for adjusting the image position by a user or aligning an image in such an apparatus.

Further, in order to realize the positioning device of the present invention by a computer such as a generally available personal computer, for example, the functions of the positioning device 400 shown in FIG. 1 are created by a program that can be interpreted by a computer. Then, the created program may be recorded on a recording medium, and the computer that realizes the functions of the present invention may read the recording medium on which the program is recorded.

[0081]

A first effect of the present invention is that, with the positioning apparatus according to the present invention, the variation between images including the amount of translation and rotation can be obtained with high accuracy. This eliminates blocks that are likely to give erroneous block matching processing results, classifies the block matching processing results into groups that are considered to have originated from the same position parameter, and optimizes the parameters obtained for each pair. This is the effect of the configuration in which the parameters are determined.

A second effect of the present invention is that, with the positioning apparatus of the present invention, the fluctuation between images including the amount of translation and rotation can be obtained at high speed. this is,
This is an effect of a configuration in which block matching processing is applied only to a block that is likely to give a correct block matching processing result.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration example of a positioning device according to the present invention.

FIG. 2 is a diagram for explaining a processing flow in the embodiment of the positioning device according to the present invention.

FIG. 3 is a diagram for explaining terms in an embodiment of the positioning apparatus according to the present invention.

FIG. 4 is a diagram for explaining processing in a block selecting unit in the embodiment of the positioning device according to the present invention.

FIG. 5 is a diagram for explaining processing in a block selection unit in the embodiment of the positioning device according to the present invention.

FIG. 6 is a diagram for describing processing in a block matching unit in the embodiment of the positioning device according to the present invention.

FIG. 7 is a diagram for explaining processing in a block classification unit in the embodiment of the positioning device according to the present invention.

FIG. 8 is a diagram for explaining processing in a block classification unit in the embodiment of the positioning device according to the present invention.

FIG. 9 is a diagram illustrating an example of a configuration of a composite image generation device using the image position adjustment device and the position alignment device according to the present invention.

FIG. 10 is a diagram for explaining an example of a configuration of a position correction device which is a component of a composite image generation device using an image position adjustment device and a position alignment device according to the present invention.

[Explanation of symbols]

 400 Positioning device 401 Block dividing means 402 Block selecting means 403 Block matching means 404 Matching result classifying means 405 Position correction parameter calculating means 406 Optimal position correction parameter determining means 407 Block attribute memory 408 Position parameter memory 450 Reference image coordinate system 451 Reference image 452 target image 453 block i 454 center of block i 455 corresponding point of block i 456 search range 460 target image coordinate system 461 block A 462 block B

Claims (6)

(57) [Claims] 1. A positioning device for determining a positional relationship between a reference image and a target image by performing a block matching process, and performing positioning between the reference image and the target image. From the result of the matching process, a matching result classifying unit that classifies blocks considered to produce similar position correction parameters into the same group, and a position correction parameter of the target image is obtained for each group classified by the matching result classifying unit. Position correction parameter calculating means; and optimal position correction parameter determining means for determining a position correction parameter that gives the best result by referring to the reference image and the target image from the obtained plurality of position correction parameters. A positioning device comprising at least: 2. The alignment device according to claim 1, further comprising: a block selection unit that selects a block that is unlikely to cause an erroneous result when performing the alignment. And performing a block matching process using only the blocks selected in (1). 3. A positioning device for determining a positional relationship between a reference image and a target image stored in an image memory and performing positioning between the reference image and the target image, wherein the reference image is stored in a plurality of positions. A block dividing unit configured to divide the block into blocks and store attribute information of each block in a first memory; and by referring to the attribute information of each block and the image memory stored in the first memory, Determine the characteristic value of each block, and a block selection unit that selects a block that is less likely to produce an erroneous result when performing alignment, and for the block selected by the block selection unit,
A block matching unit that performs a positioning process based on parallel movement, determines a point corresponding to the reference image and the target image for each of the selected blocks, and stores the result in the first memory; A matching result classifying unit that classifies blocks that are supposed to produce similar position correction parameters into the same group based on the alignment result of each block obtained by the block matching unit; and for each group classified by the matching result classifying unit. A position correction parameter calculating unit that calculates a position correction parameter of the target image and stores the position correction parameter in the second memory; and a position correction parameter for each group stored in the second memory. Optimal position correction parameter to select the position correction parameter that gives the best overlap state Aligning apparatus characterized by comprising: a determining means. 4. An optimal position correction parameter determining means,
By referring to the reference image and the target image,
The positioning apparatus according to claim 3, wherein a position correction parameter is selected. 5. A computer readable recording a positioning program for positioning the reference image and the target image by obtaining a positional relationship between the reference image and the target image by performing a block matching process. A recording medium, a matching result classifying function of classifying blocks that are likely to produce similar position correction parameters from the result of the block matching process for each block into the same group, and a group classified by the matching result classifying unit. A position correction parameter calculation function for obtaining a position correction parameter of the target image for each of the plurality of obtained position correction parameters; and a position correction parameter that gives the best result by referring to the reference image and the target image from a plurality of obtained position correction parameters. And a function for determining an optimum position correction parameter to be determined. A computer-readable recording medium having recorded thereon an alignment program. 6. A computer readable by a computer storing a positioning program for determining a positional relationship between a reference image and a target image stored in an image memory provided in the computer and positioning the reference image and the target image. A block dividing function of dividing the reference image into a plurality of blocks, and storing attribute information of each block in a first memory provided in a computer; and each of the plurality of blocks stored in the first memory. A block selection function for determining a characteristic value of each block by referring to the attribute information of the block and the image memory, and selecting a block that is less likely to cause an erroneous result when performing alignment; and For the block selected by the selection means,
A block matching function of performing a positioning process based on parallel movement, determining a point corresponding to the reference image and the target image for each of the selected blocks, and storing the result in the first memory; A matching result classifying function of classifying blocks that are likely to generate similar position correction parameters into the same group based on the alignment result of each block obtained by the block matching unit; and for each group classified by the matching result classifying unit. A position correction parameter calculation function for obtaining a position correction parameter of the target image and storing the position correction parameter in a second memory provided in a computer; and a position correction parameter for each group stored in the second memory. Select the position correction parameter that gives the best overlap of the target image Suitable position correction parameter determination function and, recording a computer-readable recording medium alignment program characterized in that it comprises a.